AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Correlation-Weighted Multi-Reward Optimization for Compositional Generation

Jungmyung Wi, Hyunsoo Kim, Donghyun Kim

ECCV 2026 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add correlation-based reweighting (CR) (ablation)CMO SD3.5-M · ConceptMix avg full markavg full mark: 0.3531 -> 0.3913Improved
add correlation-based reweighting (CR) (ablation)CMO SD3.5-M · ConceptMix avg concept fractionavg concept fraction: 0.7993 -> 0.8126Improved
add single-attribute training prompts (ablation)CMO SD3.5-M · ConceptMix avg full markavg full mark: 0.2713 -> 0.3305Improved
add multi-concept training prompts (ablation)CMO SD3.5-M · ConceptMix avg full markavg full mark: 0.3305 -> 0.3531Improved
apply DRO (MixGRPO + GDPO) alone on standard data vs base (ablation)SD3.5-M · ConceptMix avg full markavg full mark: 0.2667 -> 0.2713Mixed
increase group size from 8 to 16 (ablation)CMO SD3.5-M · ConceptMix avg full markavg full mark: 0.2934 -> 0.3919Improved
group size 4 instead of 16 (ablation)CMO SD3.5-M · ConceptMix avg full markavg full mark: 0.3919 -> 0.0000Hurt
train FLUX.1-dev with CMOFLUX.1-dev · ConceptMix full mark (k=7)full mark score: 0.0372 -> 0.1883Improved
CMO vs Qwen-Image at k=7Qwen-Image · ConceptMix full mark (k=7)full mark score: 0.1747 -> 0.1883Improved
CMO vs Qwen-Image on concept fraction at k=7Qwen-Image · ConceptMix concept fraction (k=7)concept fraction score: 0.7942 -> 0.7884Hurt
train SD3.5 with CMOSD3.5 · T2I-CompBenchAvg score: 0.5239 -> 0.6141Improved
train FLUX.1-dev with CMOFLUX.1-dev · T2I-CompBenchAvg score: 0.4866 -> 0.5961Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy